Downloaded from https://iranpaper.

ir
https://www.tarjomano.com/order

International Journal of Thermal Sciences 49 (2010) 2191e2200

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

journal homepage: www.elsevier.com/locate/ijts

Confined flow and heat transfer across a triangular cylinder in a channel

S. Srikanth, A.K. Dhiman*, S. Bijjam
Department of Chemical Engineering, Indian Institute of Technology, Roorkee 247 667, India

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, fluid flow and heat transfer across a long equilateral triangular cylinder placed in a hori-
Received 25 January 2010 zontal channel is studied for Reynolds number range 1e80 (in the steps of 5) and Prandtl number of 0.71
Received in revised form for a fixed blockage ratio of 0.25. The governing Navier-Stokes and energy equations along with
11 June 2010
appropriate boundary conditions are solved by using a commercial CFD solver FLUENT (6.3). The
Accepted 14 June 2010
computational grid is created in a commercial grid generator GAMBIT. The flow and temperature fields
Available online 16 July 2010
are presented by stream-line and isotherm profiles, respectively. The wake/recirculation length, mean
drag coefficient and average Nusselt number, etc. are calculated for the above range of conditions studied
Keywords:
Triangular cylinder
here. The critical value of the Reynolds number (i.e., transition to transient) is found to lie between
Blockage Re ¼ 58 and Re ¼ 59. The average Nusselt number and the wake length increase with increasing value of
Wake length the Reynolds number; however, the mean drag coefficient decreases with increasing value of the Rey-
Drag nolds number. Finally, simple correlations for wake length, mean drag coefficient and average Nusselt
Nusselt number number are obtained for the range of conditions studied here.
Streamlines Ó 2010 Elsevier Masson SAS. All rights reserved.
Isotherms

1. Introduction observed at Re ¼ 45 for a blockage ratio of 0.25. For symmetric flow,

the presence of triangular prism has only local weak effects on the
When placed in a fluid stream, some obstacles generate sepa- heat transfer and the flow pattern, but in periodic flow an
rated flow over a substantial portion of their surface and hence can augmentation in the value of average Nusselt number is observed
be classified as bluff bodies. On sharp edged cylinders (e.g., cylin- with increasing value of the Reynolds number. Subsequently,
ders of triangular and rectangular cross-sections, etc.), the location Abbassi et al. [2] carried out mixed convection of air in a horizontal
of separation is fixed due to fixed separation points, whereas on channel (b ¼ 0.25) with a built-in triangular prism for Grashof
cylinders with continuous surface curvature (e.g., cylinders of number ranging from 0 to 1.5  104 at Pr ¼ 0.71 for a fixed Reynolds
circular and elliptical cross-sections, etc.) the location of separation number of 100. The von Karman street favours the formation of
depends both on the shape of the body and the state of the convective cells and also the presence of the triangular prism leads
boundary layer. At low Reynolds numbers, the flow around the bluff to an increase in more than 44% of the time-averaged Nusselt
body is stable, but as the Reynolds number is increased a critical number. Zielinska and Wesfreid [3] numerically investigated the
value is reached beyond which instabilities develop. Flow past bluff wake flow behind an equilateral triangular obstacle for the Rey-
bodies has been investigated by many researchers both numerically nolds number range 34e50 with a fixed blockage ratio of 1/15. They
and experimentally because of the variety of engineering applica- found a critical Reynolds number of 38.3, which was further
tions such as electronic cooling, heat exchanger systems, offshore confirmed by the experiments of Goujon-Durand et al. [4]. Recently,
structures and flow metering devices, etc. The present work is De and Dalal [5] examined the laminar flow and heat transfer past
concerned with the flow and heat transfer across a confined equi- a triangular cylinder placed in a horizontal channel for the Reynolds
lateral triangular cylinder. number range 80  Re  200 and blockage ratio range 1=12  b 
Abbassi et al. [1] studied the two-dimensional forced convection 1=3 by using the finite volume method. A criss-cross motion of the
of air (Pr ¼ 0.71) in a horizontal channel differentially heated with vorticity layers has been observed at b ¼ 1=3, whereas they remain
a built in triangular prism for two Reynolds numbers of 30 and 100 in their own half for other cases. The SteRe curve has a flat
in detail. The transition from symmetric flow to periodic flow is maximum around Re ¼ 130 for b < 1=6, whereas it increases
monotonically for b  1=6 with Reynolds number and for Re > 130
with b. The local Nusselt number changes with time only in the rear
* Corresponding author. Tel.: þ91 1332 285890 (off), þ91 9410329605 (mob). end of the cylinder due to the vortex shedding, whereas in other
E-mail address: [email protected] (A.K. Dhiman). two faces it remains unchanged.

1290-0729/$ e see front matter Ó 2010 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ijthermalsci.2010.06.010
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2192 S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200

Nomenclature TN temperature of the fluid at the inlet, K

*
Tw constant wall temperature at the surface of the
b side of an equilateral triangular cylinder, m cylinder, K
cp specific heat of the fluid, J kg1 K1 t time (¼t * =ðb=Umax Þ)
CD drag coefficient (¼2FD =rUmax2 b) Umax maximum velocity at the channel inlet, m s1
CLrms rms value of the lift coefficient Vx component of velocity in the x-direction (¼Vx* =Umax )
f vortex shedding frequency, s1 Vy component of velocity in the y-direction (¼Vy* =Umax )
FD drag force per unit length of the cylinder, N m1 x stream-wise coordinate (¼x* =b)
h local heat transfer coefficient, W m2 K1 Xd downstream distance of the cylinder, m
h average heat transfer coefficient, W m2 K1 Xu upstream distance of the cylinder, m
H height of the computational domain, m y transverse coordinate (¼y* =b)
k thermal conductivity of the fluid, W m1 K1
L length of the computational domain, m Greek symbols
Lr wake length, m b blockage ratio (¼b/H)
Nu local Nusselt number (¼hb/k) q temperature (¼ðT *  TN Þ=ðTw*  T Þ)
N
1 1
Nu average Nusselt number (¼hb=k) m viscosity of the fluid, kg m s
p pressure (¼p* =ðrUmax
2 Þ) r density of the fluid, kg m3
Pr Prandtl number (¼mcp =k)
Re Reynolds number (¼rUmax b=m) Superscript
St Strouhal number (¼fb=Umax ) * dimensional variable

On the other hand, sufficient information is available in the physical properties) are given by Eqs. (1)e(4). In the present work,
literature on the flow around a triangular cylinder in the turbulent thermo-physical properties of the streaming fluid are assumed to
flow regime (El-Sherbiny [6]; Nakagawa [7]; Tatsuno et al. [8]; be independent of the temperature and the viscous dissipation
Wahed et al. [9]; Sharif and Gu [10]; Camarri et al. [11]; Chatto- effects are also neglected such that the present results are appli-
padhyay [12]; Peng et al. [13] and Srigrarom and Koh [14]). cable to situations where the temperature difference is not too
Thus, as far as known to us, it can be summarized here that only large and for moderate viscosity and/or shearing levels. For
Abbassi et al. [1] carried out limited study on the flow and heat instance, in this study, the temperature difference between the
transfer across a triangular cylinder in the steady confined flow streaming fluid and the triangular cylinder is assumed to be small
regime for a blockage ratio of 0.25. It is also worth mentioning here *  T Þ ¼ 2 K) so that the variation of physical properties
(i.e., ðTw N
that they used a very coarse grid size of 105  21 having first grid (e.g., density, viscosity, thermal conductivity, etc.) with tempera-
points at distances of 0.18 and 0.2 in the x- and y-directions, ture could be neglected.
respectively. In the unsteady (periodic) regime, De and Dalal [5]
examined the flow and heat transfer past a confined triangular Continuity equation
cylinder for the Reynolds number range 80  Re  200 and
blockage ratio range 1=12  b  1=3. Hence, the objective of this
vVx vVy
study is set to investigate the flow and heat transfer across þ ¼ 0 (1)
a triangular cylinder for the range of conditions: Re ¼ 1e80, vx vy
Pr ¼ 0.71 and b ¼ 0:25. The numerical investigations have also
been carried out to determine the value of the critical Reynolds x-Momentum equation
number (i.e., transition to transient). Finally, simple correlations for
wake length, mean drag coefficient and average Nusselt number are
 !
provided in order to calculate the intermediate values of wake vVx vðVx Vx Þ v Vy Vx vp 1 v2 Vx v2 Vx
þ þ ¼  þ þ (2)
length, mean drag coefficient and average Nusselt number for the vt vx vy vx Re vx2 vy2
above range of conditions.

2. Geometrical configuration and governing equations

In this study, the 2-D confined flow of an incompressible fluid in

Equilateral Triangular cylinder,
a channel with a built-in long equilateral triangular cylinder is
V x = 0, V y = 0, Tw
investigated, as shown in Fig. 1. This obstacle of triangular cross-
section is exposed to a parabolic velocity field with maximum
velocity, Umax and uniform temperature, TN at the inlet. The b H
U max
triangular cylinder is located in the middle at an upstream distance
of Xu =b from the inlet and at a downstream distance of Xd =b from
the outlet. The total length of the computational domain is L/b in
the axial direction; however, the height of the computational
domain is H/b in the lateral direction. The blockage ratio here is y
defined as b ¼ b=H. x
For the 2-D incompressible flow, the non-dimensional forms of Xu Xd
L
the continuity, the x- and y- components of momentum and energy
equations (assuming negligible dissipation and constant thermo- Fig. 1. Schematics of the flow around a confined triangular cylinder.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200 2193

y-Momentum equation walls of the channel, respectively and having 100 grid points on
each side of the equilateral triangular cylinder. The percentage
relative changes in the values of the mean drag coefficient and the


 !
vVy v Vx Vy v Vy Vy vp 1 v2 Vy v2 Vy mean Nusselt number for the above two grids are found to be less
þ þ ¼  þ þ (3) than 0.5% and less than 0.1%, respectively for Re ¼ 100. The corre-
vt vx vy vy Re vx2 vy2
sponding changes in the values of the mean drag coefficient and the
mean Nusselt number are found to be less than 1.2% and less than
Energy equation
0.05%, respectively for Re ¼ 150. Thus, the grid size of 94,794 cells is
used in this work.

 !
vq vðVx qÞ v Vy q 1 v2 q v2 q The effects of upstream distance on the values of the physical
þ þ ¼ þ (4) parameters are carried out for Xu ¼ 10b; 12b, 14b and 40b for the
vt vx vy Re Pr vx2 vy2
lowest value of the Reynolds number of unity. The relative
percentage changes in the values of mean drag coefficients are
In above equations, Reynolds number and Prandtl number are found to be about 0.0005%, 8.05  105, less than 0.005% for
defined as bUmax r=m and mcp =k, respectively. Xu ¼ 10b, 12b, 14b as compared to the value of drag coefficient
The dimensionless boundary conditions may be written as forXu ¼ 40b, respectively. However, the corresponding differences
follows (Fig. 1). in the values of the mean Nusselt number are found to be less than
 2 10.20%, 5.19%, 5.0% for Xu ¼ 10b, 12b, 14b as compared to the value
 
 At the inlet boundary, Vx ¼ 1  1  2y H ð0  y  HÞ; of the mean Nusselt number for Xu ¼ 40b, respectively. Thus, the
Vy ¼ 0; q ¼ 0. upstream distance of 12b is used here.
 On upper and lower boundary, Vx ¼ 0; Vy ¼ 0; vq=vy ¼ 0. Similarly, the effects of downstream distance are carried out for
 On the surface of the triangular cylinder, Vx ¼ 0; Vy ¼ 0; Xd ¼ 18b; 20b and 22b for the value of the Reynolds number of 80.
q ¼ 1. The percentage changes in the values of mean drag coefficients are
 At the exit boundary, the default outflow boundary condition in found to be about 0.5% and less than 0.05% for Xd ¼ 20b and 22b
FLUENT, which assumes a zero diffusion flux for all flow vari- w.r.t. the value of the mean drag coefficient for Xd ¼ 18b, respec-
ables, is used. This is similar to Neumann boundary condition tively. The corresponding relative differences in the values of mean
as vVx =vx ¼ 0; vVy =vx ¼ 0; vq=vx ¼ 0. Nusselt number are found to be less than 1.2% for Xd ¼ 20b and
Xd ¼ 22b as compared to the value of the mean Nusselt number for
Xd ¼ 18b. Thus, the downstream distance of 20b is used here.
3. Grid structure and numerical methodology In the present work, finite volume method is used on a collo-
cated grid using SIMPLE (semi implicit method for pressure linked
The grid structure in the vicinity of the equilateral triangular equations) algorithm. The second order upwind scheme is used to
cylinder is shown in Fig. 2. The grid is generated by using GAMBIT, discretize convective term, while the diffusive term is discretized
a pre-processor of FLUENT (6.3) and has both uniform and non- by central difference scheme. The resulting algebraic equations are
uniform grid distributions having a fine grid in the region of larger solved by GausseSiedel iterative scheme. The residuals of the
gradients and coarse grid in the region of low gradients. The continuity, x- and y- components and energy equations are used of
computational grid consists of 94,794 cells and each side of the the order of 1010 in the steady state regime and of 1020 in the
equilateral triangular cylinder has 100 grid points with a very fine unsteady state regime.
grid size of 0.002 units is used near the cylinder and near the top
and bottom walls of the channel. 4. Results and discussion
For the grid independence study, two grids of grid sizes of
74,660 cells and 94,794 cells are tested with smallest grid sizes of In this study, fluid flow and heat transfer across a long equilat-
0.004b and 0.002b near the cylinder and near the top and bottom eral triangular cylinder confined in a horizontal channel is studied
for the following range of conditions: Re ¼ 1e80 (in the steps of 5),
Pr ¼ 0.71 and b ¼ 0.25. In addition, numerical simulations have also
been carried out in order to investigate the critical Reynolds
number (i.e., transition to transient).

4.1. Validation of results

The benchmarking of the present results is done with the results

of De and Dalal [5] for the values of the Reynolds numbers of 80, 100
and 150 in Table 1. An excellent agreement is found between the
present results and that of De and Dalal [5]. For instance, the
maximum relative changes in the values of the mean drag coeffi-
cient and the average Nusselt number are found to be about 0.26%
and about 5.86%, respectively for the Reynolds number of 80 (Table 1).
The maximum deviations in the values of the mean drag coefficient
and the average Nusselt number are found to be about 0.55% and
less than 5%, respectively for the Reynolds number of 100. However,
the corresponding deviations in the values of the mean drag coef-
ficient and the average Nusselt number are found to be less than
1.2% and about 4.8%, respectively for the Reynolds number of 150.
The maximum changes in the values of Strouhal number are found
Fig. 2. Non-uniform grid structure (magnified view) for the blockage ratio of 0.25. to be less than 1.6%, less than 2.25% and less than 1.7% for Reynolds
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2194 S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200

Table 1 numbers of 80, 100 and 150, respectively. The larger deviations in
Validation of present results with literature values. the values of the Nusselt numbers between the present results and
Source CD St Nu that of De and Dalal [5] are probably due to the fact that De and
Re ¼ 80 Dalal (2007) used a coarser grid as opposed to the grid used in the
Present 1.6357 0.1919 4.8695 present study. For instance, De and Dalal [5] used a grid size of
De and Dalal [5] 1.64 0.195 4.6 28,000 cells with having 70 control volumes on each side of the
Re ¼ 100 triangular cylinder, the smallest and the largest grid sizes used are
Present 1.6708 0.2004 5.5624 0.004b and 0.5b, respectively; however, in the present study, the
De and Dalal [5] 1.68 0.205 5.3 grid consists of 94,794 cells with having 100 control volumes on
Re ¼ 150 each side of the cylinder, the smallest and the largest grid sizes used
Present 1.9349 0.2212 7.1270 are 0.002b and 0.4b, respectively. Further, the open reported liter-
De and Dalal [5] 1.96 0.225 6.8
ature suggests that the maximum changes upto about 5% are
acceptable [1,2,15,16]. Thus, based upon the information available

4 4
a Re = 1 e Re = 30
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
b Re = 5 f Re = 40
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
c Re = 10 g Re = 50
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
d Re = 20 h Re = 55
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

Fig. 3. Streamline profiles for Re ¼ 1, 5, 10, 20, 30, 40, 50 and 55.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200 2195

in the literature and based upon my own experience, the present suggest that the critical Reynolds number exits between the Rey-
results are believed to be reliable less than 5%. This validates the nolds numbers of 58 and 59 for the blockage ratio of 0.25 (Fig. 4).
present numerical solution procedure. Unfortunately, no experi- However, Abbassi et al. [1] reported that the flow is changing from
mental result is available in the literature for the range of condi- symmetric to periodic at Re ¼ 45 for a blockage ratio of 0.25. This
tions studied here. discrepancy is due to the fact that Abbassi et al. [1] used a very
coarse grid size of 105  21 having first grid points at distances of
4.2. Flow patterns 0.18 units and 0.2 units in the x- and y-directions, respectively as
opposed to the very fine grid size used in the present study. For
The two-dimensional flow patterns in the vicinity of the long instance, in the present study, the grid size consists of 94,794 cells
equilateral triangular obstacle are presented by streamline having the smallest grid size of 0.002 units near the cylinder and
contours for 1  Re  55 in the steady flow regime in Fig. 3(aeh). near the top and bottom walls of the channel, respectively.
Similar to the long square cylinder case, it is found that no flow Furthermore, in this study, it is observed that flow becomes
separation occurs from the surface of the triangular cylinder at unsteady (periodic) for Re  59 for the blockage ratio of 0.25. Fig. 5
a Reynolds number of unity as the viscous force dominates the flow (aed) presents the instantaneous streamline profiles for the high-
here (Fig. 3a). As the Reynolds number gradually increased, flow est value of the Reynolds number of 80 used in this work and show
separates from the rear edge of the triangular obstacle and forms the periodic variations in the wake behind the obstacle.
two symmetric vortices behind the triangular cylinder above and
below the mid plane. The size of these vortices increases with 4.3. Isotherm patterns
increasing value of the Reynolds number for 1 < Re  55 and
b ¼ 0.25. On further increasing the value of the Reynolds number The representative steady isotherm profiles around the trian-
(i.e., Re > 55), at one instance flow becomes erratic and symmetric gular obstacle for Re ¼ 1e55 for the blockage ratio of 0.25 are
behaviour disappears, von Karman vortex street triggers. Fig. 4 shown in Fig. 6(aeh). Clearly the top and bottom surfaces have the
(aed) shows the temporal history of the lift coefficient for the maximum crowding of the temperature contours as compared to
Reynolds numbers of 58 and 59 (Fig. 4a, b) and the corresponding the rear surface of the triangular cylinder. This is due to the
streamline profiles (Fig. 4c, d) for the blockage ratio of 0.25. It can increased circulation of large amount of fluid with increasing
be seen from these figures that the value of the lift coefficient Reynolds number, and also with an increase in the value of the
decays with time very slowly at a Reynolds number of 58 and is Reynolds number, the turning of isotherms on the rear surface can
likely to reach a steady state after a long time; however, for the also be seen. For the range of conditions studied here, flow field
Reynolds number of 59, the value of the lift coefficient increases becomes unsteady periodic for 59  Re  80 and the instantaneous
with time and the wake loses its steady state. These findings isotherms for the extreme value of the Reynolds number of 80 are

4
a Re = 58 c Re = 58
3E-06 3.5

2E-06 3

1E-06 2.5
CL

0 2
y

-1E-06 1.5

-2E-06 1

-3E-06 0.5

0
1100 1200 1300 1400 1500 1600 11 12 13 14

0.025 b Re = 59 4 d Re = 59
0.02 3.5
0.015
3
0.01
0.005 2.5
CL

0 2
y

-0.005 1.5
-0.01
1
-0.015
-0.02 0.5

0
0 1000 2000 3000 11 12 13 14
t x
Fig. 4. Temporal lift coefficient and streamline profiles at Re ¼ 58 and 59.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2196 S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200

shown in Fig. 5(eeh). It can be seen here that a temperature street values of the drag coefficient are utilized to calculate the time-
is formed behind the confined triangular cylinder, which is very averaged values of the drag. Similarly, the temporal variations of
similar to Karman vortex street. the lift coefficient and the Nusselt number with time are shown in
Fig. 7cef, respectively for the Reynolds numbers of 60 and 80.
Similar to the time-averaged drag coefficient, the time-averaged
4.4. Time history values of the lift and Nusselt number are also calculated here by
using the stabilized values of the lift and Nusselt number.
The temporal variation of the drag coefficient in the unsteady
(periodic) flow regime for the values of the Reynolds numbers of 60
and 80 is presented in Fig. 7a, b. The instantaneous values of the 4.5. Wake length
drag coefficient are calculated at each time step and plotted versus
time. It is found that the value of the drag initially oscillates and The variation of the wake length with Reynolds number in
finally stabilizes with time forRe  59. Further, these stabilized the steady regime is shown in Fig. 8 (a). This figure also includes

4 4
a t=T e t=T
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
b t = T/4 f t = T/4
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
c t = 2T/4 g t = 2T/4
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
d t = 3T/4 h t = 3T/4
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

Fig. 5. Instantaneous streamline and temperature profiles for Re ¼ 80.

 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200 2197

the variation of wake length with Reynolds number in case of larger in the case of the triangular cylinder than that of the
a confined square cylinder for the blockage ratio of 0.25 (Dhiman square cylinder case in the steady confined flow regime for the
et al. [15,16]). Since both square and equilateral triangular blockage ratio of 0.25.
cylinders (of side b) have the same projected area and this fact In order to calculate the wake length for the triangular cylinder
has been utilised here in order to compare the present results case for the intermediate values of the Reynolds number, the
with that of the square cylinder. Similar to the square cylinder following simple expression can be used for the range 10  Re  55
case, the wake length behind the triangular cylinder increases and b ¼ 0:25.
monotonically with increasing value of the Reynolds number in
the steady confined flow regime. It is due to the fact that the Lr =b ¼ 0:21 þ 0:15Re0:70 (5)
triangular cylinder has fixed separation points that lead to the
wake length and the size of the vortex is increasing predomi- Equation (5) has a maximum error of less than 1.20% with the
nantly with Reynolds number. However, the wake length is computed present results of the wake length.

4 4
a Re = 1 e Re = 30
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
b Re = 5 f Re = 40
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
c Re = 10 g Re = 50
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

4 4
d Re = 20 h Re = 55
3 3

2 2

1 1

0 0
11 12 13 14 15 16 11 12 13 14 15 16

Fig. 6. Isotherm contours for Re ¼ 1, 5, 10, 20, 30, 40, 50 and 55.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2198 S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200

a Re = 60 c Re = 60 e Re = 60
1.7009 0.04
4.16024
0.03

1.7007 0.02 4.16022

CD 0.01
4.1602

Nu
CL
1.7005 0
4.16018
-0.01

1.7003 -0.02 4.16016

-0.03
4.16014
1.7001 -0.04
1425 1450 1475 1500 1050 1075 1100 1125 1150 1175 1200 1425 1450 1475 1500
4.8705

1.644 0.1
b Re = 80 d Re = 80 f Re = 80
1.64 4.8699
0.05
CD
1.636

Nu
CL

0
4.8693
1.632
-0.05

1.628
-0.1 4.8687
1.624
175 200 225 225 250 275 300 325 350 375 175 200 225
t t t

Fig. 7. Temporal variation of drag coefficient, lift coefficient and Nusselt number for Re ¼ 60 and 80.

a b

c d

Fig. 8. Variation of (a) wake length, (b) friction drag, (c) pressure drag and (d) total drag coefficients with Reynolds number.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200 2199

4.6. Drag coefficient

0.12 a
The drag is the force exerted by the fluid on the bluff body in the
direction on the flow. There are two components contribute to 0.1
drag: friction drag coefficient (CDF ) and pressure drag coefficient
(CDP ). Therefore, the total drag coefficient can be written as 0.08
CD ¼ CDF þ CDP . Fig. 8(bed) presents the variation of the individual CLrms
and overall drag coefficients with Reynolds number for a fixed 0.06
blockage ratio of 0.25. The individual and overall drag values for the
long square cylinder case are also presented in these figures (Dhi-
0.04
man et al. [15,16]). It is observed that the value of the friction drag,
pressure drag and mean drag coefficients decreases with increasing
value of the Reynolds number for both the obstacles of triangular 0.02
and square cross-sections. At low Reynolds numbers, drag is
strongly depend on the Reynolds number because of the viscous 0
60 65 70 75 80
forces play a dominant role in the steady flow regime. From these
figures (Fig. 8bed), it can be observed that the values of the pres-
sure drag coefficient and the overall drag coefficient for the square
cylinder case are higher than that of the triangular cylinder case.
0.192 b
However, the values of the friction drag coefficient for the square 0.19
cylinder case are smaller for 10  Re  55 and higher for Re < 10
than that of the triangular cylinder case. 0.188
In this study, the time-averaged drag coefficient is calculated by St
averaging at least 10 cycles beyond the time the asymptotic shed- 0.186
ding frequency of Karman vortex is attained. It is found that
the value of the time-averaged drag coefficient decreases 0.184
with increasing value of the Reynolds number for 59  Re  80 and
b ¼ 0:25 (Fig. 8d). 0.182
The simple drag correlation is obtained for the range of condi-
tions studied here. The mean drag coefficient for the intermediate 0.18
values of the Reynolds can be calculated by using Eq. (6) for the
range 1  Re  80 and b ¼ 0:25. 60 65 70 75 80
Re
CD ¼ 1:23 þ 36:02Re1:04 (6)
Fig. 9. Variation of rms value of lift coefficient and Strouhal number as a function of
Reynolds number.
The above expression has a maximum error of less than 2.4% with
the computed present results.
Further, the rms value of the drag coefficient is calculated for the
Nusselt number is obtained by averaging the averaged Nusselt
Reynolds number range 59  Re  80. It is found that the rms value
number for each surface of the triangular obstacle.
of the drag coefficient increases from 7.18  105 to 6.804  103
The variation of the average Nusselt number with Reynolds
with Reynolds number (59  Re  80).
number for the values of the Reynolds numbers of 1 to 80 for the
blockage ratio of 0.25 is presented in Fig. 10. In order to present the
4.7. Lift coefficient and Strouhal number comparison of the triangular cylinder average Nusselt number with
that of the square cylinder average Nusselt number, the variation of
The time-averaged lift coefficient is calculated by averaging at the square cylinder average Nusselt number is also shown in Fig. 10
least 10 cycles (beyond the time the asymptotic shedding frequency
of Karman vortex is attained). The time-averaged lift coefficient
remains approximately zero (in the order of about 105e106) for
59  Re  80 for the blockage ratio of 0.25. The variation of rms
values of the lift coefficient with Reynolds number is also presented
in Fig. 9(a). The rms value of the lift coefficient increases as the
value of the Reynolds number increase in the unsteady periodic
regime.
Since the flow field becomes unsteady periodic for Re  59 for
the fixed value of the blockage ratio of 0.25, the variation of
Strouhal number with Reynolds number for 59  Re  80 is pre-
sented in Fig. 9(b). It can be seen that Strouhal number increases
with increasing value of the Reynolds number for the above range
of conditions.

4.8. Average Nusselt number

In this study, the average Nusselt number is obtained by aver-

aging the local Nusselt number over the triangular cylinder
surfaces. The overall mean value of the triangular cylinder average Fig. 10. Variation of average Nusselt number with Reynolds Number.
 Downloaded from https://iranpaper.ir
https://www.tarjomano.com/order

2200 S. Srikanth et al. / International Journal of Thermal Sciences 49 (2010) 2191e2200

(1  Re  45) (Dhiman et al. [15,16]). The average cylinder Nusselt found to be about 25% for Re ¼ 1 and 12.5%e15% for 5  Re  45
number increases monotonically with increasing value of the w.r.t. the values of the triangular cylinder case. Finally, the simple
Reynolds number in both the cases. However, the average Nusselt correlations of wake length, mean drag coefficient and average
number is found to be higher in the case of triangular cylinder than Nusselt number are obtained for the range of conditions studied.
that of the square cylinder. The maximum percentage relative
difference between the values of the average Nusselt number for
triangular and square obstacles is found to be about 25% for Re ¼ 1 References
and 12.5%e15% for 5  Re  45 w.r.t. the values of the triangular
cylinder average Nusselt number. [1] H. Abbassi, S. Turki, S.B. Nasrallah, Numerical investigation of forced convec-
The triangular cylinder average Nusselt number for the inter- tion in a plane channel with a built-in triangular prism, Int. J. Thermal Sci. 40
(2001) 649e658.
mediate values of the Reynolds numbers can be calculated by using [2] H. Abbassi, S. Turki, S.B. Nasrallah, Mixed convection in a plane channel with
Eq. (7) for the range of conditions: 10  Re  80, Pr ¼ 0.71 and a built-in triangular prism, Numer. Heat Transfer, Part A 39 (2001) 307e320.
b ¼ 0:25. [3] B.J.A. Zielinska, J.E. Wesfreid, On the spatial structure of global modes in wake
flow, Phys. Fluids 7 (1995) 1418e1424.
[4] S. Goujon-Durand, P. Jenffer, J.E. Wesfreid, Downstream evolution of the
Nu ¼ 1:32 þ 0:17ðRe PrÞ0:74 (7) Benard von Karman instability, Phys. Rev. E 50 (1994) 308e313.
[5] A.K. De, A. Dalal, Numerical study of laminar forced convection fluid flow and
This simple correlation has a maximum error of less than 1.5% for heat transfer from a triangular cylinder placed in a channel, J. Heat Transfer
10 < Re  80 and about 2.0% for Re ¼ 10 with the computed present 129 (2007) 646e656.
[6] S. El-Sherbiny, Flow separation and reattachment over the sides of a 90
results.
triangular prism, J. Wind Eng. Ind. Aerodyn. 11 (1983) 393e403.
The rms value of the Nusselt number is also calculated here and [7] T. Nakagawa, Vortex shedding mechanism from a triangular prism in
found that the rms Nusselt number increases with increasing value a subsonic flow, Fluid Dyn. Res. 5 (1989) 69e81.
of the Reynolds number. The rms value of the Nusselt number is [8] M. Tatsuno, T. Takayama, M. Tatsuno, M. Tatsuno, On the stable posture of
a triangular or a square cylinder about its central axis in a uniform flow, Fluid
found to increase from 1.25  105 to 6.5  104 with Reynolds Dyn. Res. 6 (1990) 201e207.
number for the range 59  Re  80. [9] A.K. El-Wahed, M.W. Johnson, J.L. Sproston, Numerical study of vortex shed-
ding from different shaped bluff bodies, Flow Meas. Instrum. 4 (1993)
233e240.
5. Conclusions [10] M.A.R. Sharif, Z. Gu, Turbulent flow simulation behind a V-shaped flame
stabilizer using a nonlinear ke3 model and a smoothing algorithm, Chem. Eng.
In the present study, 2-D incompressible flow and heat transfer Commun. 189 (2002) 471e488.
[11] S. Camarri, M.V. Salvetti, G. Buresti, Large-eddy simulation of the flow around
around a long confined equilateral triangular cylinder has been a triangular prism with moderate aspect ratio, J. Wind Eng. Ind. Aerodyn. 94
examined for Reynolds number ¼ 1e80 and Prandtl number of 0.71 (2006) 309e322.
for a fixed blockage ratio of 0.25. The streamline and isotherm [12] H. Chattopadhyay, Augmentation of heat transfer in a channel using a trian-
gular prism, Int. J. Thermal Sci. 46 (2007) 501e505.
contours are presented in order to describe the flow and temper-
[13] J. Peng, X. Fu, Y. Chen, Experimental investigations of Strouhal number for
ature fields near the triangular obstacle, respectively. The critical flows past dual triangulate bluff bodies, Flow Meas. Instrum. 19 (2008)
value of the Reynolds number is found to exist between Re ¼ 58 and 350e357.
[14] S. Srigrarom, A.K.G. Koh, Flow field of self-excited rotationally oscillating
Re ¼ 59. The mean drag coefficient decreases with increasing value
equilateral triangular cylinder, J. Fluids Struct. 24 (2008) 750e755.
of the Reynolds number; however, the wake length increases with [15] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Flow and heat transfer across
Reynolds number for the range of conditions covered here. The a confined square cylinder in the steady flow regime: effect of Peclet number,
average Nusselt number increases with increasing value of Int. J. Heat Mass Transfer 48 (2005) 4598e4614.
[16] A.K. Dhiman, R.P. Chhabra, V. Eswaran, Steady flow across a confined square
the Reynolds number. The maximum change between the values of cylinder: effects of power-law index and blockage ratio, J. Non-Newtonian
the average Nusselt number for triangular and square obstacles is Fluid Mech. 148 (2008) 141e150.